Privileged scaffolds for blocking protein-protein interactions: 1,4-Disubstituted naphthalene antagonists of transcription factor complex HOX-PBX/DNA

Article abstract of DOI:10.1016/j.bmcl.2004.05.068

Structure-based-design studies, with the crystal structure of the HOXB1-PBX1/DNA transcription factor complex, were used to identify 1,4-disubstituted naphthalenes as potential antagonists. An initial library of 32 analogs was synthesized, two of which were found to be more potent than the reported activity for a 12 amino acid peptide antagonist. Antagonists were also identified of the related BRN1/DNA and BRN2/DNA transcription factor complexes indicating that a 1,4-disubsituted naphthalene may be a privileged scaffold for preparing screening libraries targeting this family of transcription factor complexes.

Full text of DOI:10.1016/j.bmcl.2004.05.068

Bioorganic & Medicinal ChemistryLetters 14 (2004) 3875–3879
Privileged scaffolds for blocking protein–protein interactions:
1,4-disubstituted naphthalene antagonists of transcription
factor complex HOX–PBX/DNA
Tao Ji,a, Madison Lee,b, Steven C. Pruittb and David G. Hangauera,*
aDepartment of Chemistry, University at Buffalo, The State University of New York, Buffalo, NY 14260–3000, USA
bDepartment of Molecular and Cellular Biology, Roswell Park Cancer Institute, Buffalo, NY 14263, USA
Received 24 February2004; revised 26 May2004; accepted 27 May2004
Available online 19 June 2004
Abstract—Structure-based-design studies, with the crystal structure of the HOXB1–PBX1/DNA transcription factor complex, were
used to identify1,4-disubstituted naphthalenes as potential antagonists. An initial libraryof 32 analogs was synthesized, two of
which were found to be more potent than the reported activityfor a 12 amino acid peptide antagonist. Antagonists were also
identified of the related BRN1/DNA and BRN2/DNA transcription factor complexes indicating that a 1,4-disubsituted naphthalene
maybe a privileged scaffold for preparing screening libraries targeting this familyof transcription factor complexes.
ꢀ 2004 Elsevier Ltd. All rights reserved.
The term ‘privileged structure’ was first introduced by
Evans et al.1 This term has been used to denote a
molecular framework (scaffold) that is able to provide
ligands for individual multiple targets when suitably
decorated with side chains. A varietyof privileged
scaffolds have been reported since that time.2 Focused
combinatorial libraries, based upon these privileged
scaffolds, provide enriched hit rates when screening for
receptor agonists/antagonists, ion channels modulators,
or enzyme inhibitors. In contrast to these more tradi-
tional targets, discovering small molecules capable of
disrupting protein–protein interactions has proven to be
more challenging, partlybecause the contact surface
between bound proteins typically covers a relative large
surface area.3 Random high-throughput screening for
antagonists of protein–protein interactions has generally
been less successful than has screening against the tra-
ditional categories of drug targets. When the generally
accepted MW 500 limit for compounds likelyto be
orallyactive is applied to the reported antagonists of
protein–protein interactions, then the majorityof the
remaining antagonists have potencies in the 1–100 lM
range. Most of these antagonists were discovered by
4
screening, with onlya minorityobtained bydesign.
Transcription factor complexes are responsible for much
of the regulation of gene expression5 and therefore offer
manypotential protein–protein interaction drug tar-
gets,6 including anticancer targets.7 The 39 members of
the HOX family8a and 4 members of the PBX family8b
are proteins that bind DNA as heterodimers to form
transcription factor complexes. The numerous hetero-
dimeric HOX/PBX combinations playcritical and
complex roles in transcriptional regulation during pat-
9
terning in earlyembryonic development and manyare
utilized again in specific tissues of the adult.10 A variety
11
of cancers displayaltered HOX gene regulation.
For
example, in leukemias and lung cancers HOX genes are
overexpressed and consequentlyantagonists of the
HOX/PBX complexes controlling these genes offers the
potential for novel anticancer drugs. Also small mole-
cule antagonists of individual (or subsets) HOX/PBX
complexes can be used as pharmacological tools to
investigation their function.
Herein we report the first small molecule antagonists of
HOX–PBX protein–protein interactions.12 A crystal
structure of the minimal HOXB1 and PBX1 fragments
necessaryfor cooperative DNA binding has been
reported.13 This structure showed that HOXB1 and
PBX1 bind to overlapping sites on opposite faces of the
Keywords: Protein–protein interaction antagonists; Transcription fac-
tor complex antagonists; HOX; PBX; BRN.
* Corresponding author. Tel.: +1-716-645-6299; fax: +1-716-645-6963;
e-mail: hangauer@acsu.buffalo.edu
Contributed equallyto this paper.
0960-894X/$ - see front matter ꢀ 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2004.05.068

3876
T. Ji et al. / Bioorg. Med. Chem. Lett. 14 (2004) 3875–3879
DNA double helix. However, the two proteins also bind
to each other, and therebystabilize the ternarycomplex,
where HOXB1 places a short peptide arm (FDWMK)
into a hydrophobic binding pocket (a ‘hot spot’) on the
surface of PBX1. Previous in vitro studies have shown
that disruption of the cooperative DNA binding by
HOX to PBX1 proteins can be accomplished bypoint
mutations in this HOX peptide domain.14 These studies
also showed that the Trp 182 (W) and Met 183 (M)
residues within the highlyconserved pentapeptide region
of the HOX familyof peptide domains (F/Y-P/D- W-M-
K/R) were critical for cooperative DNA binding with
PBX. The crystal structure shows HOXB1 Trp 182 and
Met 183 packing against each other and binding in the
hydrophobic pocket on the surface of PBX1.13
This cooperative binding can be competitivelyblocked
Figure 1. As shown, the naphthalene ring is positioned
within the hydrophobic pocket, the C-1 amide side chain
NH is forming a hydrogen bond with Leu-252 and the 3-
fluoro atom on the phenyl amide side chain is hydrogen
bonding with Try-291. The C-4 ether amide side chain is
extending across the surface of PBX1 in the direction of
the bound DNA. From these modeling studies the C-4
side chain appeared to have multiple potential binding
opportunities so a range of side chains at this position
was selected for experimental evaluation (Table 1).
A focused combinatorial libraryof thirt-ytwo 1,4-
disubstituted naphthalene derivatives with the modeled
3-fluorophenyl amide C-1 side chain and various ether
amide side chains at C-4 were synthesized (Fig. 1). The
synthesis begins from commercially available 4-meth-
oxy-1-naphthaldehyde 1 as outlined in Figure 1. Oxi-
dation of aldehyde 1 provided acid 2.17 Amide 3 was
obtained bycoupling the corresponding acid chloride
with 3-fluoroaniline. Demethylation of methyl ether 3
using AlCl3/EtSH produced the corresponding phenol,18
which was then converted to methyl ester 4. Acid 5 was
obtained byrefluxing ester 4 with LiOH in methanol.
The final target amides 6 were synthesized using PyBOP
as the coupling reagent with a broad range of amines.
All final products were purified bysilica gel chroma-
tographyto greater than 95% purityand gave 1H NMR
and MS spectra that are consistent with the expected
product.
bythe 12-residue peptide, QPQI YPWMRKLH (IC50
100 lM), containing the conserved pentapeptide
¼
sequence.14
The human HOXB1–PBX1/DNA (code 1B72) complex
described above was downloaded from the Protein Data
Bank (www.rcsb.org). All molecular modeling, ligand
docking, and librarydesign studies were carried out
using SYBYL 6.8, and the associated modules FlexX,
CScore, and LeapFrog, all obtained from Tripos, Inc.
(www.tripos.com). The pentapeptide region of HOXB1
that binds in the hydrophobic pocket on the surface of
PBX1 as described above was deleted to expose the
critical peptide-binding region for designing small mol-
ecule antagonists, and the resulting complex was mini-
The selectivityof individual members of the libraryfor
the HOXA1–PBX1/DNA target, as well as the abilityof
some analogs to cross-over to other transcription factor
targets was evaluated bytesting them in parallel against
the BRN1 and BRN2 transcription factors. BRN1 and
BRN2 are members of mammalian class I POU tran-
scription factor familyand are expressed in the devel-
oping embryonic brain.19 POU domain class
transcription factors contain, in addition to the POU
domain, a homeodomain helix-turn-helix class DNA
binding motif similar to that found in the HOX and
PBX class transcription factors.20 EMSAs were per-
formed with BRN1 and BRN2 exactlyas was done with
HOXA1–PBX1 except that a sequence recognition ele-
ment known to bind BRN1/BRN2 was used as the
probe.21
mized.
A
varietyof potential scaffolds, with
representative appended side chains, were docked into
the now exposed hydrophobic pocket and the sur-
rounding surface of PBX1. These candidate scaffolds
were selected bya combination of visual evaluation of
the binding surface, hand docking of candidate ligands,
automated docking (FlexX), de novo design experiments
(LeapFrog), ease of synthesis, and predictions of binding
affinity(FlexX and CScore). An important criteria that
was also applied to the candidate scaffold selection pro-
cess is the abilityto produce potential antagonists with
MW<500 and that have the abilityto meet the additional
15
‘rule of 5’ criteria developed byLipinski et al.
for
compounds likelyto be successful as oral therapeutics.
Finally, rigid scaffolds that result in antagonists with a
limited number of rotatable bonds were given priority
since this rigidityis also predicted to improve the prob-
Of the 32 analogs listed in Table 1, 24 showed measur-
able inhibition of the formation of the HOX1–PBX1/
DNA ternarycomplex when screened at a 300 lM initial
concentration. A relativelyhigh initial screening con-
centration was chosen for two reasons: (1) the known
12-residue peptide antagonist, QPQIYPWMRKLH,
containing the conserved pentapeptide sequence
(underlined) that binds to the Pbx1 hydrophobic pocket,
has an IC50 of only100 lM, (2) protein–protein inter-
action targets are significantlymore challenging to block
with small molecules than the classical drug targets for
which lower screening concentrations can be used.
16
abilityof obtaining orallyactive drugs.
With the above criteria in mind the first scaffold selected
for synthesis and testing was the 1,4-disubstituted
naphthalene scaffold (Table 1). This rigid scaffold pro-
vides two diversityside chains able to interact with the
PBX protein at the mouth of the hydrophobic pocket
and a phenyl ring to penetrate into the hydrophobic
pocket. The libraryof potential antagonists prepared
from this scaffold all have molecular weights between
340 (7) and 489 Da (36) and have a limited number of
freelyrotating bonds (Table 1).
Of these 24 active compounds against the HOXA1–
PBX1/DNA target the most potent were 30 and 31 with
an IC50’s ¼ 86 and 65 lM, respectively. These com-
A modeled complex (after minimization) of the parent
inhibitor based upon this scaffold, 7, is illustrated in

T. Ji et al. / Bioorg. Med. Chem. Lett. 14 (2004) 3875–3879
3877
a
Table 1. HOXA1–PBX1/DNA, BRN1/DNA, and BRN2/DNA antagonist activityof 1,4-disubsituted naphthalenes
OCH CONR
2
R
2
1
4
1
O
N
H
F
6
Compound
R1; R2
HOXA1–PBX1 IC50 (lM)b
BRN1 IC50 (lM)c
BRN2 IC50 (lM)c
7
H; H
NH2; H
272
Low activity(300 lM)
8
9
Methyl; H
Ethyl; H
Low activity (300 lM)
200
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
n-Propyl; H
Isopropyl; H
278
272
Isobutyl; H
tert-Butyl; H
298
No activity (300 lM)
173
Cyclopropyl; H
Propargyl; H
810
162
>10,000
204
331
111
2,2,2-Trifluoroethyl; H
2-Hydroxyethyl; H
2-Benzyloxyethyl; H
2-Dimethylaminoethyl
–CH2COOCH2CH3; H
–CH2COOH; H
Phenyl; H
No activity (300 lM)
Low activity (300 lM)
No activity (300 lM)
No activity(300 lM)
No activity(300 lM)
254
433
2-Fluorophenyl; H
3-Fluorophenyl; H
4-Fluorophenyl; H
2-Methoxyphenyl; H
3-Methoxyphenyl; H
4-Methoxyphenyl; H
3-Hydroxyphenyl; H
4-Hydroxyphenyl; H
3,5-Dimethoxyphenyl; H
2,5-Dimethoxyphenyl; H
3,4-Dimethoxyphenyl; H
2,4-Dimethylphenyl; H
3,5-Dimethoxybenzyl; H
Methyl; methyl
Low activity (300 lM)
293
718
580
161
86
142
124
148
138
65
Low activity (300 lM)
Low activity (300 lM)
464
Low activity (300lM)
No activity (300 lM)
No activity (300 lM)
No activity(300 lM)
R1, R2 ¼ (CH2CH2)2O
a Determined byelectrophoretic mobilityshift assay(EMSA). IC
data results from an averaged densitometric analysis of multiple EMSA gels.
50
b 2.5 ng each of the mouse HOXA1 192–288 protein and the mouse/human 233–319 PBX1 protein in a final volume of 25 lL binding buffer con-
taining 10 mM HEPES, pH 7.9, 1 mM EDTA, 1 mM DTT, 5% Ficoll-400, 0.1 mg/mL BSA, and 135 mM NaCl was pre-incubated for 15 min at 4 ꢁC
with 0.5 lg polydIdC (to reduce nonspecific DNA binding) and 1 lL of an appropriate dilution of the inhibitorycompound in DMSO. Proteins in
binding buffer were incubated for 30 min at approximately22 ꢁC with 1 lL of 32P labeled oligonucleotide probe prepared as described below (1–
2 · 104 CPM) prior to electrophoresis on a 4% polyacrylimide gel in 50 mM TBE buffer at 150 V for 2 h 15 min. Two oligonucleotides, HOXPBX-CT
Top, 50CTCTCCTTTTGATTGATTAA-30, and HOXPBX-CT Bottom 50-AGAGCTTAATCAATCAAAAGG-30 were annealed and the ends
extended with Klenow in the presence of a32P-dCTP to prepare the oligonucleotide probes. Following electrophoresis, gels were dried and exposed
to X-rayautoradiographic film for 18–24 h.
c EMSAs were performed with BRN1 and BRN2 exactlyas was done with HOXA1–PBX1 when using nuclear extracts from RA treated P19 cells (for
BRN induced byRA in P19 cells see Ref. 23) except that a sequence recognition element known to bind BRN1 or BRN2 monomers was used as the
probe.21
pounds are more potent than that reported for the much
larger 12-residue peptide antagonist QPQIYPWMR-
KLH containing the conserved pentapeptide sequence16
and differ from each other onlybythe position of the
hydroxyl group in the C-4 aryl amide moiety.
indicated that these substituents are directed toward the
bound DNA. Consequently, the larger 4-hydroxyphenyl
group maybe altering the conformation of the PBX1/
DNA binarycomplex so as to destabilize it whereas the
smaller cyclopropyl group does not extend far enough
into the DNA binding region to do this. On the other
hand both antagonists can bind in the hydrophobic
pocket therebypreventing HOXA1 from binding to the
binarycomplex. Upon closer evaluation of the 31 dose–
response (data not shown) of the binaryPBX/DNA
band intensityrelative to the ternaryHOXA1–PBX/
DNA band intensityit was noted that the binaryPBX/
Figure 2 illustrates that 31 inhibits the formation of the
PBX1/DNA binarycomplex as well the ternaryHOX1–
PBX1/DNA complex whereas 15 inhibits formation of
ternarycomplex only. Antagonist 31 differs from 15 by
replacing the small cyclopropyl R1 with the larger 4-
hydroxyphenyl group. Our modeled complex (Fig. 1)

3878
T. Ji et al. / Bioorg. Med. Chem. Lett. 14 (2004) 3875–3879
DNA band initially increases in intensityat low 31
concentrations. One potential explanation for this result
is that HOXA1 is released from the ternarycomplex
generating a higher initial concentration of the binary
complex. As the concentration of 31 is further increased
the binaryPBX/DNA complex formation is inhibited
enough to overcome the release of this binarycomplex
from the ternarycomplex and therefore the binaryband
intensitybegins to decrease.
Collectively, the above results suggest, but do not prove
that compounds based on the 1,4-disubstituted naph-
thalene scaffold are binding to PBX1 rather than directly
binding to DNA. If 15 were directlybinding to DNA
then it would need to be binding to a region of the DNA
not contacted byPBX1 in order to explain the selective
inhibition of the formation of the HOXA1–PBX1/DNA
ternarycomplex versus the PBX1/DNA binarycomplex.
Likewise if 31 were directlybinding to DNA then it
would need to be binding to a region of DNA not
contacted byPBX1 since low concentrations of 31 in-
crease the concentration of the PBX1/DNA binary
complex. Since both results are more readilyexplained
bybinding of the compounds in the pentapeptide
binding pocket, for which theyhave complementary
topological features bydesign, it is likelythat the
mechanism of inhibition is direct binding to PBX1.
Figure 1. Modeled complex of 7 docked into PBX1/DNA hydrophobic
pocket with hydrogen bonds indicated by dash lines. Ligand 7 is in
green, DNA is in red, PBX1 is atom color coded.
OCH3
OCH3
OCH3
a
b
O
H
O
OH
O
N
H
F
1
2
3
c, d
OCH2CONR1R2
OCH2COOH
OCH2COOCH3
4
1
f
Figure 2 illustrates that 15 is also selective toward the
HOXA1–PBX1/DNA target (IC50 ¼ 173 lM) versus
e
BRN1/DNA
(IC50 ¼ 810 lM)
or
BRN2/DNA
O
N
H
F
O
N
H
5
F
O
N
H
4
F
(IC50 ¼ >10,000 lM). This result again suggests direct
binding to PBX1 rather than to DNA. Extension 15
from the pentapeptide pocket into the DNA binding
region provided 31, which now inhibits BRN1 and
BRN2 from binding to their recognition sequences with
IC50’s of 124 and 138 lM, respectively, versus 65 lM
against HOXA1–PBX1/DNA. Overall, 15 is selective for
the blocking the HOXA1–PBX1/DNA ternarycomplex
relative to all of the other complexes tested whereas 31 is
an active antagonist of all of the complexes with some
selectivityfor the HOXA1–PBX1/DNA complex. Since
31 can extend into the DNA binding region of the
PBX1/DNA complex, and therebyblock DNA from
binding, BRN1 and 2 mayhave similar hydrophobic
pockets resulting in the abilityof 31 to block DNA from
binding to them as well.
6
Scheme 1. (a) NaClO2, NaH2PO4, H2O, t-BuOH, 2-methyl-2-butene,
rt, 24 h, 95%; (b) (1) SOCl2, CH2Cl2, reflux (50 ꢁC), 6 h; (2) 3-fluoro-
aniline, THF, DIEA, reflux, 0.5 h, 90%; (c) AlCl3, CH2CH2SH,
CH2Cl2, 0 ꢁC to rt, 2 h, 93%; (d) BrCH2COOCH3, K2CO3, THF,
reflux, 1 h, 100%; (e) LiOH, MeOH, H2O, reflux, 2 h, 95%; (f) PyBOP,
R1R2NH, DMF, 0 ꢁC to rt, 2 h, 90–95%.
The abilityof the 1,4-disubstituted naphthalenes to
cross-over from PBX to the BRN transcription factors
and antagonize their DNA complex, when suitable side
chains are present, suggests that this maybe a privileged
scaffold for related transcription factors. The known
tendency22 of proteins to reuse general binding motifs
for a new purpose, after evolving the necessarymodifi-
cations to attain a new function, suggests that the PBX
hydrophobic pocket targeted by these antagonists may
be reused in BRN and other transcription factor com-
plexes. Consequently, larger screening libraries based
upon this scaffold mayproduce antagonists, including
more potent ones, of various members of this class of
transcription factors. A libraryof such antagonists,
including those selective for ternaryversus binary
Figure 2. Selectivityof 15 and 31 for HOXA1–PBX1/DNA, PBX1/
DNA, BRN1/DNA, and BRN2/DNA.

T. Ji et al. / Bioorg. Med. Chem. Lett. 14 (2004) 3875–3879
3879
7. Karamouzis, M. V.; Gorgoulis, V. G.; Papavassiliou, A.
G. Clin. Cancer Res. 2002, 8, 949.
8. (a) Scott, M. P. Cell 1992, 71, 551; (b) Knoepfler, P. S.;
Kamps, M. P. Mech. Dev. 1997, 63, 5.
9. Gehring, W. J.; Affolter, M.; Burglin, T. Annu. Rev.
Biochem. 1994, 63, 487.
complexes as discovered herein, would serve as valuable
bio-probes (or drug leads) for investigating the role of
these transcription factors in various biological pro-
cesses and diseases.
10. van Oostveen, J.; Bijl, J.; Raaphorst, F.; Walboomers, J.;
Meijer, C. Leukemia 1999, 13, 1675.
Acknowledgements
11. Maroulakou, I. G.; Spyropoulos, D. D. Anticancer Res.
2003, 23, 2101.
12. Ji, T.; Lee, M.; Pruitt, S. C.; Hangauer, D. G. Patents
pending through Universityat Buffalo.
13. Piper, D. E.; Batchelor, A. H.; Chang, C. P.; Cleary, M.
L.; Wolberger, C. Cell 1999, 96, 587.
14. Knoepfler, P. S.; Kamps, M. P. Mol. Cell. Biol. 1995, 15,
5811.
This work was supported in part byIRCAF and STOR/
UB Product Development Fund grants, both from the
Universityat Buffalo to D.G.H. and a grant from the
NIH (RO1HD36416) to S.C.P. The authors thank Drs.
George DeTita and TimothyUmland for the gift of
purified HOXA1 and PBX1, and AmyBussman, Cath-
erine Franz, and Elzbieta Kawinski for technical assis-
tance.
15. Lipinksi, C. A.; Lombardo, F.; Dominy, B. W.; Feeney,
P. J. Adv. Drug Deliv. Rev. 1997, 23, 3.
16. Veber, C. F.; Johnson, S. R.; Cheng, H.-Y.; Smith, B. R.;
Ward, K. W.; Kopple, K. D. J. Med. Chem. 2002, 45,
2615.
References and notes
17. Ahmed, A.; Bragg, R. A.; Clayden, J.; Tchabanenko, K.
Tetrahedron Lett. 2001, 42, 3407.
18. Node, M.; Nishide, K.; Fuji, K.; Fujita, E. J. Org. Chem.
1980, 45, 4275.
19. McEvilly, R. J.; de Diaz, M. O.; Sconemann, M. D.;
Hooshmand, F.; Rosenfeld, M. G. Science 2002, 295,
1528.
1. Evans, B. E.; Rittle, K. E.; Bock, M. E.; DiPardo, R. M.;
Freidinger, R. M.; Whitter, W. L.; Lundell, G. F.; Veber,
D. F.; Anderson, P. S.; Chang, R. S. L.; Lotti, V. J.;
Cerino, D. J.; Chen, T. B.; Kling, P. J.; Kunkel, K. A.;
Springer, J. P.; Hirshfield, J. J. Med. Chem. 1988, 31,
2235.
2. Horton, D. A.; Bourne, G. T.; Smythe, M. L. Chem. Rev.
2003, 103, 893.
20. (a) Josephson, R.; Muller, T.; Picker, J.; Okabe, S.;
Reynolds, K.; Turner, P. A.; Zimmer, A. Development
1998, 125, 3087; (b) Hara, Y.; Rovescalli, A. C.; Kim, Y.;
Nirenberg, M. PNAS 1992, 89, 3280.
21. Hox/Pbx and Brn binding sites mediated Pax3 expression
in vitro and in vivo. Pruitt, S. C.; Bussman, A.; Maslov, A.
Y.; Natoli, T. A.; Heinaman, R. Gene Expression Patterns
2004, in press. Available online 20 May2004.
22. Breinbauer, R.; Vetter, I. R.; Waldmann, H. Angew.
Chem., Int. Ed. 2002, 41, 2878.
3. For reviews of protein–protein interaction antagonists see:
(a) Toogood, P. L. J. Med. Chem. 2002, 45, 1543; (b)
Thorsten, B. Angew. Chem., Int. Ed. 2003, 42, 2462; (c)
Boger, D. L. Bioorg. Med. Chem. 2003, 11, 1607.
4. Boger, D. L.; Desharnais, J. S.; Capps, K. Angew. Chem.,
Int. Ed. 2003, 42, 4138.
5. Wolgerger, C. Annu. Rev. Biophys. Biomol. Struct. 1999,
28, 29.
6. Emery, J. G.; Ohlstein, E. H.; Jaye, M. Trends Pharmacol.
Sci. 2001, 22, 233.
23. Bain, G.; Ray, W. J.; Yao, M.; Gottlieb, D. I. Bioessays
1994, 16, 343.